If these proofs were output by the AI in a format readable by a proof verification system, the verification step of publishing vanishes. Then it's only valuable to check if the stated intention actually matches the proof and isn't something completely different.
I don't know why your comment is downvoted, because it makes much sense. That's assuming the proof is not founded in assumptions, and the proof checker doesn't have bugs.
Not sure about "bugs" in that area, but there is a lot of work going on by mathematicians in formalizing and checking ever more complex proofs using proof assistants.
These systems have been tested on very complex proofs already, but well... I'm not a mathematician, just a software engineer who accepted his new role in this "new thinking order".